Conceptual and Methodological Issues in Consumer Satisfaction Research

ABSTRACT - Consumer satisfaction has recently become a particularly salient concern for both business and government. This increased interest, however, has generated a relatively meager amount of research designed to assess the determinants of consumer satisfaction and dissatisfaction. This paper critically analyzes the existing survey and experimental research in this area, focusing on conceptual and methodological issues. An alternative model of consumer satisfaction based upon Thibaut and Kelley's (1959) comparison level theory is proposed.

Citation:

Stephen A. LaTour and Nancy C. Peat (1979) ,"Conceptual and Methodological Issues in Consumer Satisfaction Research", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 431-437.

Advances in Consumer Research Volume 6, 1979      Pages 431-437

CONCEPTUAL AND METHODOLOGICAL ISSUES IN CONSUMER SATISFACTION RESEARCH

Stephen A. LaTour, University of Florida

Nancy C. Peat (student), University of Florida

[The authors wish to express their appreciation to Olli Ahtola and Pauline Houlden for their helpful comments and suggestions.]

ABSTRACT -

Consumer satisfaction has recently become a particularly salient concern for both business and government. This increased interest, however, has generated a relatively meager amount of research designed to assess the determinants of consumer satisfaction and dissatisfaction. This paper critically analyzes the existing survey and experimental research in this area, focusing on conceptual and methodological issues. An alternative model of consumer satisfaction based upon Thibaut and Kelley's (1959) comparison level theory is proposed.

INTRODUCTION

Consumer satisfaction has long been a central concern of modern marketing practitioners (Kotler, 1976) and more recently a major concern of various governmental agencies engaged in consumer protection activities. The consumer-ist movement has made consumer satisfaction an even more salient concern for both business and government by calling attention to consumer dissatisfaction with products, services, and marketing practices.

Unfortunately, this increased interest in and concern about consumer satisfaction/dissatisfaction has generated a relatively meager amount of research designed to unravel the determinants of consumer satisfaction/dissatisfaction. The research can be classified into two major streams: 1) survey research designed to assess a) incidences of satisfaction/dissatisfaction for various products and services (e.g. Handy and Pfaff, 1975), and b) these incidences together with an assessment of the relationship between satisfaction and various socioeconomic and demographic characteristics (e.g. Handy, 1977; Ash, 1978) and 2) mostly experimental research designed to assess the impact of expectations about product performance on post-consumption satisfaction or perceived product performance (e.g. Cardozo, 1965; Olshavsky and Miller, 1972; Anderson, 1973). The purpose of this paper is to critically evaluate the conceptual and methodological problems of this research and to propose an alternative conceptualization of the determinants of consumer satisfaction based upon Thibaut and Kelley's (1959) comparison level theory.

CRITICAL REVIEW OF CONSUMER SATISFACTION RESEARCH

Survey Research

Survey research has revealed that degree of consumer satisfaction/dissatisfaction varies widely among individuals as well as over product and service categories. Some survey researchers have attempted to account for differences in satisfaction by correlating consumer satisfaction/dissatisfaction with an extensive variety of socioeconomic and demographic variables. The usefulness of this "shotgun approach" (c.f. Kassarjian, 1971) is questionable, not only because it capitalizes on chance, but also because it lacks a clearly stated conceptual rationale. In addition, the statistically significant relationships that have been obtained usually account for relatively small percentages of variance in consumer satisfaction (c.f. Gr°nhaug, 1977; Day and Bodur, 1977; Ash, 1978). Thus while survey research has provided some useful descriptive data, it has not contributed very much to our understanding of the determinants of consumer satisfaction/dissatisfaction.

Research on Consumer Expectations

Some consumer researchers have theorized that satisfaction is a function of the discrepancy between a consumer's expectations about the performance of a product and obtained product performance. A number of functional relationships between expectations and obtained performance have been proposed to explain satisfaction, but the research has concentrated on three basic approaches. [Other functional relationships such as those based upon the theory of generalized negativity (Carlsmith and Aronson, 1963) have been suggested, but these generally have not been replicated and will not be discussed.] First is Howard and Sheth's (1969), Engel, Kollat and Blackwell's (1973), and Cardozo's (1965) contrast hypothesis. "Contrast theory," traceable to Helson's (1964) adaptation-level theory, suggests that consumers will compare actual performance of a product to their expectations about performance. If obtained performance is less than expected, consumers will be dissatisfied. On the other hand, if their expectations are met or exceeded, they will be satisfied.

The second approach is derived from assimilation-contrast theory (Sherif and Hovland, 1961). According to this theory, expectations serve as an anchor for the judgment of product performance, and assimilation or contrast effects will occur as a function of the degree of discrepancy between obtained and expected performance (c.f. Anderson, 1973, Olson and Dover, 1976). If the discrepancy is not too large, assimilation should occur. In this event expectations higher than obtained outcomes should result in higher performance judgments (and therefore greater satisfaction) than if expectations match obtained performance. Similarly, expectations lower than obtained outcomes should result in perceptions of poorer performance (and less satisfaction) than if expectations match obtained outcomes. If, however, the discrepancy is very large, a contrast effect should occur; expectations higher (lower) than obtained outcomes should result in lower (higher) performance judgments (and consequently, less (more) satisfaction) than if expectations match obtained outcomes. This theory may be viewed as a refinement of "contrast theory," for contrast effects are hypothesized to occur only when the discrepancy between expected and obtained performance is extreme.

The third approach, based upon various psychological consistency theories (c.f. Olshavsky and Miller, 1972; Anderson, 1973), hypothesizes that consumers will experience psychological tension when an inconsistency exists between an expected and obtained level of performance. The unpleasantness generated by this tension will cause consumers to (mis)perceive product performance as consistent with their expectations. Thus as expectations increase, perceived product performance should increase and greater satisfaction should result.

A few investigations have been conducted to test the validity of these models of consumer satisfaction. Unfortunately, rather than measuring the impact of expectations on satisfaction, they have usually measured the impact of expectations on perceived product performance or quality (Cardozo, 1965; Anderson, 1973; other studies such as those by Olshavsky and Miller (1972) and Olson and Dover (1976) have been misclassified by various researchers as studies of the relationship between expectations and satisfaction when in fact they only claim to study the relationship between expectations and perceived performance). The investigators have simply assumed that assessment of perceived product performance on quality is equivalent to assessing satisfaction; however, perceived product performance or quality is not necessarily the same as satisfaction. While perceived product performance is likely to be highly correlated with amount of satisfaction, there are other variables that may interfere with this relationship. For example, a product or service provider may perform well on a number of dimensions but the product or service may cost too much or be too distant from the consumer. In such instances the consumer could be dissatisfied. In addition, unless performance measures contain an affective component (e.g. on this dimension the product performs poorly, neither poorly nor well, well), they will not indicate whether the consumer will be satisfied or dissatisfied with the obtained level of performance. For example, Anderson (1973) measured perceived quality by having subjects estimate the price of a pen they evaluated. There is no method of determining from this data what level of quality would be required for the consumer to describe himself or herself as satisfied rather than dissatisfied. In sum, care must be taken in making inferences about satisfaction on the basis of quality or performance judgments. Nonetheless, a careful examination of the research concerning the effect of expectations on perceived performance is in order because the amount of variance in satisfaction accounted for by product quality or performance is likely to be fairly high. Unfortunately, examination of this research reveals serious methodological problems which cause further difficulties in making inferences about the effect of expectations on consumer satisfaction.

For example, Cardozo (1965) found contrast effects when subjects were asked to evaluate desirability and quality of ballpoint pens. But closer examination of his results indicates that anchors for the dependent measures varied as a function of whether expectations were high or low (c.f. Olshavsky and Miller, 1972). Thus the validity of Cardozo's contrast effect may be questioned and the effect of expectations on satisfaction is not clarified by this study.

Olshavsky and Miller (1972) found that higher expectations produced higher performance judgments and claimed support for a consistency theory interpretation of the results. Analysis of their methodology suggests, however, that this interpretation is untenable. Olshavsky and Miller simply asked subjects to evaluate a product presented to them. In this circumstance, degree of involvement with the task would have been extremely low since subjects made no choice among products and their outcomes were not affected by the quality of the product as they would have been in a true purchase situation. Thus it is unlikely that the tension necessary to create consistency effects would have been established. The results might be interpreted as an assimilation effect (which does not postulate the tension required by consistency theories); however, further examination of their methodology indicates that even this interpretation is questionable. In their study different expectations were created about a tape recorder's performance by varying the ratings contained in an authoritative test report. After listening to the tape recorder described in the test report, subjects were asked to judge the performance of the recorder. Since subjects may have interpreted an instruction to rate the recorder as an instruction to rate the brand of recorder in general rather than the tested sample, they may have relied more heavily on beliefs created by the test reports than they would have otherwise. It may have been particularly important to rely upon the test report in order to make an accurate judgment since subjects listened to only one tape and may have been concerned about avoiding judgmental errors resulting from temporary factors such as the quality of the tape, the competency of the person making the recording, or a defect in this particular sample of the brand. In addition, the experiment may have created demand characteristics (Orne, 1962). There may have been an implicit demand to use the test report information in arriving at a judgment since subjects probably perceived that the experimenter intended them to use the supplied in- formation (i.e. the purpose of the experiment was probably transparent to the subjects). The probable existence of evaluation apprehension (Rosenberg, 1969) may have enhanced the impact of these demand characteristics. Subjects would have wanted the experimenter to evaluate their judgments favorably and therefore may have felt further compelled to use the information supplied by the experimenter. Thus possible judging of the brand rather than the tested sample, demand characteristics, and evaluation apprehension may have all tended to exaggerate the influence of the expectation-producing information on subjects' judgments of product performance. Hence the observed assimilation effect may have been artifactual.

Interpretation of the findings of Anderson's (1973) study is also difficult. Subjects were first given information which created differential expectations about the quality of a pen and were then asked to evaluate the pen. Anderson interpreted his results as demonstrating an assimilation effect because expectations above the objective quality level produced higher quality judgments than expectations matching obtained quality, while expectations below the objective level produced lower quality judgments. Anderson also perceived what he labeled a contrast effect -- when expectations were extremely high, judgments of product quality were somewhat lower than when expectations were moderately high. This, however, is not a true contrast effect since such an effect would require perceived quality to be less than objective quality. [The relative ineffectiveness of the extremely high expectation was probably due to subjects attaching less weight to information lacking in credibility.] As in Olshavsky and Miller's study, subjects may have interpreted the instructions to evaluate "this pen" as a request to evaluate "this brand of pen" rather than "this particular sample." In addition, they may have felt an implicit demand to use the expectation-producing information provided by the experimenter and the effect of this demand may have been magnified by the existence of evaluation apprehension. Again, the assimilation effect may well have been artificial.

Olson and Dover's (1976) investigation reveals a possible assimilation effect for judgments of coffee bitterness but the results are not statistically significant. The authors argue that one of their four measures of bitterness showed a significant difference between experimental and control groups but the probability level was only .10. This is marginal at best, given the lack of a multivariate test or some other method of guarding against Type I errors when multiple univariate tests are employed for conceptually similar dependent variables. In addition, while demand characteristics were eliminated in this study, subjects were still asked to rate the brand of coffee, not the particular sample they tested. They may have believed the sample was bitter (and would have been dissatisfied with it) but because they were asked to evaluate the brand, they may have relied less on the sample they tasted -- which could have been influenced by temporary factors -- than on the expectation-producing information supplied by the experimenters. Thus if one is interested in the effect of expectation-producing information on the consumer's response to a given purchase experience, even this small trend toward an assimilation effect may be artifactual.

Finally, Swan and Combs (1976) measured both expectations about and satisfaction with clothing purchases. They found that consumers who were dissatisfied with clothing purchases were more likely to have had their expectations negatively disconfirmed than were consumers who were satisfied with purchases. This seems to support the contrast hypothesis. Nonetheless, before accepting this interpretation of the effect of expectations on satisfaction, it should be noted that this study was based on recall of particularly satisfying and dissatisfying purchases. Thus there may have been some retrospective distortion of expectations. In addition, subjects were not asked to report middle-range experiences for which assimilation effects might have been found. If this had been done, assimilation-contrast theory might have been supported instead.

Given these problems, what can be said about the effect of expectations on satisfaction? We would speculate that expectations will have some influence on the consumer's post-consumption beliefs about the performance of a product on various attribute dimensions but only when the consumer is uncertain about the level of the attribute possessed by the product (i.e. the dimension is ambiguous). In such cases an assimilation effect should occur. (The attributes in the studies discussed above were not very ambiguous, and the fairly strong assimilation effects for judged performance or quality were probably the result of methodological problems --particularly demand characteristics. In Olson and Dover's (1976) study where the possibility of demand characteristics was reduced, the size of the assimilation effect was very small.) A contrast effect for post-consumption beliefs about the performance of a product on an attribute has never been demonstrated. Indeed, no matter how outlandish a claim by a manufacturer, we would not expect the consumer to think the product had less of an attribute than that judged to exist in the absence of a manufacturer's claims. When manufacturer's claims are extreme, the consumer would attach less weight to them and they would not influence the consumer's belief about an attribute as much as they would had the claims been moderately high. Thus for ambiguous dimensions, expectations above the objective performance level may increase judgments about the quality or performance of the product and therefore increase satisfaction. But, even though expectations above objective performance levels may increase judgments about the level of an attribute believed to be possessed by a product, there will still be some discrepancy between the level believed to be obtained and the expectation level (unless the attribute dimension is totally ambiguous). This discrepancy would not have occurred had expectations been veridical with the objective attribute level. According to contrast theory, this discrepancy which remains between the subjective and expected attribute levels will contribute to dissatisfaction. Thus the gain in satisfaction obtained by increasing perceived quality through creation of high expectations might be eliminated by the dissatisfaction associated with this discrepancy. Obviously, the more ambiguous the attribute, the more the consumer will be forced to rely on the expectation-producing information rather than personal judgment in forming a belief about the attribute level. The more ambiguous the attribute, the smaller the discrepancy between the subjectively obtained level and the expected level, and thus the more effective high expectations would be in producing greater satisfaction. Clearly, if dimensions are unambiguous, expectations above objective levels of performance would not produce any increase in perceived quality and the result of such expectations would be consumer dissatisfaction. Overall, we would hypothesize that the creation of expectations above objective performance or quality levels would typically result in some degree of consumer dissatisfaction.

FURTHER PROBLEMS AND ISSUES

Expectations -- the Primary Determinant of Consumer Satisfaction?

In addition to methodological and inferential problems with the research on consumer expectations, one may well question the theory that consumer satisfaction or dissatisfaction with products and services results primarily from the confirmation or disconfirmation of expectations created by manufacturers' or service providers' claims. [Expectations can also be the result of past experience with similar products or knowledge about other consumers' experiences with similar products. Later sections discuss the central role these experiences may play in determining satisfaction/dissatisfaction.] Let us begin by clearly specifying what is meant by expectations. As Olson and Dover (1976) have noted, researchers have been less than precise in defining this construct; however, as our review indicates, their manipulation of expectations reveals that researchers construe them to be pre-decisional or pre-consumption beliefs about the overall performance of the product created by manufacturers' claims or test-report information. Arguing from the perspective of Ahtola's (1975) multiattribute attitude model, Olson and Dover conclude that expectations should be defined as the consumer's beliefs about the levels of attributes possessed by a product. This is an approach which we favor because of its high degree of specificity and diagnostic usefulness. Both of these approaches fail, however, to recognize the role of the consumer's affective response to obtained attributes in determining satisfaction. For example, a consumer may be forced to purchase a brand whose attributes he or she does not evaluate favorably (e.g. the consumer may have moved to a new market area or the manufacturer may have discontinued marketing the product whose attributes the consumer liked). Even though the consumer's predecisional expectations (beliefs) about the attributes of the brand may be confirmed by post-purchase experience with the product, the consumer will still be dissatisfied with the product because of his/her unfavorable evaluation of its attributes.

The disconfirmation of expectations model would fail to account for consumer satisfaction/dissatisfaction in other circumstances as well. A new brand may be produced with attributes the consumer evaluates more favorably than the attributes of other brands presently on the market. Even if the manufacturer created expectations that were too high, the consumer would not be dissatisfied, for despite the negative disconfirmation of expectations, the brand possesses more of the desired attributes than did previous brands. This is not to deny the role that disconfirmation of expectations may play in determining consumer satisfaction. Indeed, it would be reasonable to expect that the consumer in the first example (forced to purchase an unsatisfactory product) might have been more dissatisfied had the disliked attributes been unexpected, and the consumer in the second example (purchasing a product with favorable but less-than-expected attributes) would have been more satisfied had the manufacturer not created unrealistic expectations. We would speculate, however, that the consumer's evaluation of the product's attributes may account for more of the variability in post-purchase satisfaction than would the confirmation or disconfirmation of expectations about those attributes.

Discriminant Validity of the Satisfaction Construct

A further issue for consumer satisfaction researchers is the discriminant validity of the satisfaction construct. Satisfaction with a product is simply an evaluative response to that product. (Although some researchers may be more interested in satisfaction with the purchase circumstances, that is also an evaluative response.) Yet, a different construct, attitude, has been more frequently employed by consumer behavior researchers to refer to evaluative responses to products. Given that attitude and satisfaction are both evaluative responses to products, it is not clear whether there are any substantive differences between the two. In fact, it may be more parsimonious to consider satisfaction measures as post-consumption attitude measures. This does not mean that satisfaction is an inappropriate measure of the post-decision evaluative response or that it should be supplanted by other measures used by attitude researchers. Rather, this implies that we should be careful about attaching any special significance to satisfaction as a construct when it may simply be one measure of an attitude.

If satisfaction with a product is essentially the same as attitude toward a product, one might ask why satisfaction research has not been integrated with attitude theory and research. In addition, if the two concepts are highly similar, one might ask whether theory and research under the label of attitudes might contribute to an understanding of theory and research under the label of satisfaction, and vice versa.

Perhaps one of the major reasons for lack of integration of the two areas is that attitude researchers have focused largely on pre-decisional evaluations of products while satisfaction researchers have focused largely on post-consumption evaluative responses. In addition, satisfaction researchers have focused on the confirmation/disconfirmation of expectations (beliefs) about attributes while attitude researchers (particularly those interested in multiattribute models) have focused on expectations (beliefs) about attributes, evaluations of those attributes, and in some instances the importance of those attributes in determining the consumer's attitude (see Fishbein and Ajzen (1975) for a discussion of the merit of importance weights). It is understandable that pre-decisional attitude research has not focused on the confirmation or disconfirmation of expectations, but it is not clear why satisfaction researchers have largely ignored the role of evaluations associated with obtained attributes. Perhaps it is time for satisfaction researchers to give more consideration to the role of these evaluations.

An Alternative Conceptualization of Consumer Satisfaction

Although many researchers in the area of consumer satisfaction are not fully acquainted with Thibaut and Kelley's (1959) comparison level theory, it may provide a useful basis for understanding post-consumption satisfaction/ dissatisfaction with products and services. A modification of this theory, described below, may also provide insight into the determinants of consumer satisfaction with each of a product's component attributes, some combination of which presumably determines the consumer's overall satisfaction with the product. [Miller (1977) appears to be the only satisfaction researcher who acknowledges the usefulness of their approach.]

Thibaut and Kelley are social psychologists who conceptualize the result of any interaction in terms of rewards and costs. In the consumer setting these rewards and costs might include (but would not necessarily be limited to) the attributes of the product or service (including its cost), the pleasures or perhaps difficulties associated with interpersonal interactions with salespeople or service providers, and the approving or disapproving responses of other persons to the purchased product. The average of obtained rewards and costs constitutes the outcome of an interaction. Clearly, some researchers might be interested in rewards and costs associated with the product itself while others might be more interested in the rewards and costs attendant to the circumstances surrounding the purchase experience.

According to Thibaut and Kelley, the consumer's satisfaction with the outcome of a purchase would be determined by the discrepancy between the outcome and a standard of comparison known as the Comparison Level (CL). Outcomes above the CL (positive discrepancies) will be satisfying while those below the CL (negative discrepancies) will be dissatisfying. [It should be noted that Thibaut and Kelley also consider the CL to be the deserved level of outcomes. It seems to us, however, that the deserved level would at least be minimally satisfying and would therefore be above the affectively neutral point on the outcome scale.] The comparison level itself is thought to be determined by the average of salient outcomes for the same or similar interactions that one has experienced or is aware of. [Thus the concept of comparison level is similar to Helson's (1964) concept of adaptation level except that the comparison level is a standard for determining evaluative responses while the adaptation level is usually considered to be a standard for non-evaluative judgments (e.g. the brightness of a light).] Outcomes that one is aware of but has not personally experienced would include known outcomes obtained by similar others (see Festinger's (1954) discussion of social comparison theory) and to a lesser extent outcomes one expects to obtain from the interaction (where these expectations are based on the unique characteristics of the present interaction rather than previous experience or the known outcomes of other people). Thibaut and Kelley's discussion suggests that these outcomes are weighted in the following order to determine the CL: first and most importantly, outcomes a person has directly experienced; second, outcomes of similar others; and third, special expectations created by the present interaction. Drawing a parallel to the situation of the consumer, this means that the comparison level for the outcome associated with any given product or service will be a function of experiences with similar products or services in the past, to a lesser extent the experiences of similar consumers, and to a still lesser extent outcomes promised by the manufacturer, retailer, and/or service provider.

Thibaut and Kelley's theory is particularly appealing because it explicitly recognizes that satisfaction is not an absolute phenomenon but rather a relative one. (Although the disconfirmation of expectations model also assumes that satisfaction results from a relative comparison of obtained attributes to expected attributes, as outlined above, the model is too restricted to account fully for consumer satisfaction.) Thus if consumers purchase a product with attributes they evaluate more positively than those obtained from previously purchased products, they will be satisfied (assuming outcomes of similar consumers and expectations specific to the purchase situation are held constant). For example, consumers who are used to tires that last 5,000 miles will be satisfied with tires that last 10,000 miles while consumers who are used to tires lasting 30,000 miles will be dissatisfied with those same tires.

While this approach to understanding overall satisfaction with a product or service appears to be useful, a word of caution is in order. Thibaut and Kelley argue that satisfaction is determined by averaging the rewards and costs associated with an interaction and comparing the result with the comparison level. However, Thibaut and Kelley define rewards as "the pleasures, satisfactions, and gratifications the person enjoys" (p. 12, emphasis ours). Their definition of costs employs a somewhat different terminology but suggests that costs are unpleasant or dissatisfying aspects of the interaction. If rewards are merely satisfactions and costs merely dissatisfactions, then the satisfaction associated with the obtained outcome is the average of the component rewards and costs and recourse to a comparison level is unnecessary to determine overall satisfaction with that outcome.

A Modified Comparison Level Theory of Satisfaction

While the comparison process described above may therefore be meaningless because the obtained outcome already contains the necessary affective information, extension of the concept of comparison to the attribute level may prove to be meaningful. We would propose that for each attribute there is a comparison level. This comparison level is a function of past experience with various levels of the attribute (for similar products used for similar purposes), levels of the attribute the consumer is aware that similar consumers receive, and expectations created by the manufacturer, salesperson, and/or service provider. These attribute levels are not measured in terms of rewards and costs but rather in terms of natural scale values (e.g. size, number of miles per gallon, number of compliments received for a new dress). For example, a consumer's satisfaction with the miles per gallon achieved by his/her new car would probably be related to the discrepancy between obtained MPG and a weighted average of his/her previous cars' average MPG, the average MPG his/her friends obtain from their cars, and the manufacturer's promises about the MPG he/she should obtain from the car. It can be speculated that the consumer's overall satisfaction with a car would be an additive function [It is conceivable that this could be an averaging function; however, it is not the purpose of this paper to further explore the adding-averaging controversy.] of the discrepancies from the comparison level for each salient attribute (e.g. MPG, size, trunk space, braking distance, riding comfort) of the car. In determining overall satisfaction, each discrepancy would be weighted by the importance a consumer attaches to that attribute dimension. In addition, the discrepancies should be thought of as percentages of the comparison level to eliminate scaling differences among the various attribute dimensions. (This can be accomplished by dividing the discrepancy by the comparison level.)

For the attributes where more of the attribute is always desirable (a positive, infinite ideal point, e.g. MPG, tire mileage), positive discrepancies should contribute to satisfaction while negative discrepancies should contribute to dissatisfaction. For attributes where less of that attribute is always more desirable, the effect of discrepancies will be reversed. If an attribute dimension involves a finite ideal point [Finite ideal points are likely to be complexly determined; there may be some physiological determinants, a particular attribute level may have been associated with happy or sad events (classical conditioning), and/or an attribute level may have been the one most frequently experienced (familiarity).] (e.g., the sweetness of a soft drink), discrepancies from two comparison levels must be considered in determining the satisfaction associated with that attribute: one below and one above the ideal point. The comparison level below the ideal point can be defined as a weighted average of the experienced, known, and promised levels of attributes below the ideal point while the comparison level above the ideal point will be a weighted average of experienced, known, and promised levels of attributes above the ideal point. Obtained attribute levels below the ideal point would be compared to the lower CL while obtained attribute levels above the ideal point would be compared to the higher CL. As long as an outcome is between the two comparison levels, a consumer will be satisfied with the attribute level. (Similar logic could be applied to a finite negative ideal point except that obtained attribute levels between the two comparison levels would be dissatisfying while all other obtained attribute levels would be satisfying.) For example, with respect to a positive, finite ideal point, a consumer might ideally prefer a soft drink to be "moderately sweet." If the consumer has always obtained this outcome level, then the ideal point would define a single comparison level and any level of sweetness above or below the ideal point would be at least somewhat dissatisfying. It is unusual, however, for a consumer to have experienced only the ideal level of outcomes. Thus some levels below and above the ideal point will probably have become slightly or moderately satisfying (though certainly not maximally satisfying as the ideal point would be). The average sweetness experienced, known, or promised below the ideal would determine the lower comparison level. Sweetness between this level and the ideal point would be described by the consumer as satisfying. The average sweetness experienced, known, or promised above the ideal would determine the upper comparison level. Sweetness between the ideal point and the upper comparison level would also be described as satisfying by the consumer. All other levels of sweetness would produce varying degrees of dissatisfaction, depending upon their discrepancy from the appropriate comparison level. Overall satisfaction with a soft drink should be a sum of the discrepancies of all relevant attributes from their appropriate CL's with each discrepancy weighted by the importance of the attribute with which it is associated.

The proposed model can be formally expressed as:

EQUATIONS  (1)  and  (2)

where:

S_j = post-consumption satisfaction with brand j

I_i = importance the consumer attaches to attribute i

"_i = subjectively experienced attribute level

CL_i = comparison level for attribute i

P_i = average of personally experienced attribute levels

R_i = average of attribute levels experienced by referent persons (similar others)

E_i = expected value of the attribute level based on the unique characteristics of the present purchase situation (i.e. expectations orthogonal to P_i and R_i such as expectations created by manufacturer's claims)

n = number of salient attributes

w₁, w₂, w₃ = empirically determined weights

It should be noted that this formula assumes a positive, infinite ideal point for each attribute. For attributes having a negative, infinite ideal point, the sign of the discrepancy is the same as in the formula for outcomes below the ideal point; but for outcomes above the ideal point, the sign must be reversed. For negative, finite ideal points, the signs of the obtained discrepancies would be reversed from what they would be for a positive, finite ideal point. Of course, the lower or upper CL should be employed depending upon whether the obtained level of an attribute is below or above the ideal point. While the formula assumes constant weighting of P_i, R_i, and E_i across the attributes, this assumption may need revision as the model is further explored. Finally, note that because the model assumes Ai is the subjectively experienced attribute level, it can account for situations in which the consumer is uncertain about the obtained level (i.e. where the dimension includes some ambiguity). In such cases the subjectively experienced attribute level is defined as the expected value of the attribute level. That is, A_i would be an additive function of the subjective probabilities associated with each attribute level multiplied by the respective values of those levels. [A similar model could be developed to determine consumer satisfaction with the purchase circumstances by specifying the salient attributes of the purchase situation. In many circumstances it may be useful to incorporate both product attribute and purchase circumstance attributes into the model.]

Implications of the Proposed Satisfaction Model for Attitude Theory

As indicated previously, attitude theory (in particular multiattribute theory) has implications for consumer satisfaction research because it reveals the necessity of considering the consumer's evaluative (satisfaction/ dissatisfaction) response to obtained attributes. Discussion is now in order concerning the implications that the proposed model of satisfaction may have for attitude theory. In particular, the extension of Thibaut and Kelley's concepts to the level of the component attributes of a product provides unique insight into the determinants of the evaluative responses to those attributes. Attitude theorists (particularly those employing Fishbein's (1967) model or variants of his model) have typically assumed that the evaluation of an attribute level is the result of classical and/or operant conditioning. It has been demonstrated that this can happen (c.f. Fishbein and Ajzen, 1975), but whether this is the predominant method by which affect is attached to various attribute levels may be questioned. We would argue that attitudes toward various attribute levels are rarely classically conditioned. It is unusual for an attribute level to be associated consistently with some particular good or bad experience (although ideal points may in part be established by such experiences). In fact, if classical conditioning of responses to attribute levels were to occur very often, affect curves would probably be very jagged -- but this is a rare occurrence. Operant conditioning would also fail as a complete explanation of affective responses to attribute levels. For example, we could say that the less time a vacuum cleaner took to clean a carpet, the more a consumer would be reinforced by using it. But saying this begs the question concerning the point at which the evaluation of amount of time spent vacuuming changes from unfavorable to favorable (i.e. from dissatisfaction to satisfaction). We would argue instead that whether the amount of time is evaluated favorably will be a function of the amount of time one has taken in the past, the amount of time one is aware that similar others' vacuum cleaners take, and expectations created by the vacuum cleaner's manufacturer. Thus the process of judgment relative to a comparison level would seem to better account for the formation of attitudes toward attribute levels.

Cautions Concerning the Proposed Model

Four final caveats are in order concerning the proposed model of satisfaction. First, consumers will not demonstrate infinite adaptive capabilities. Even if a consumer has experienced only perfumes with the pungent odor of rotten eggs, it is unlikely that such a consumer could ever be satisfied with such a perfume although he/ she would probably be less dissatisfied with such a pungent smell than most consumers in our society. In sum there are probably physiological absolutes which have an impact upon affective responses to attributes. Nonetheless most product attribute levels are probably well within human adaptive abilities.

Second, it should be noted that the proposed model assumes the attributes to be measurable on a scale with metric properties. This requirement may appear to limit the usefulness of the model when an attribute dimension is of a qualitative nature. For example, with respect to applesauce, a consumer may be concerned about whether the container is metal or glass. These may appear to be qualitatively different attributes, but this simply means that metal and glass packages differ on more than one underlying dimension. The consumer has simply used the glass-metal distinction to summarize his/her concerns about these underlying dimensions. For example, in this case the underlying dimensions might be breakage resistance and amount of alteration in taste caused by the package. These dimensions are quantifiable and employable in the proposed model.

Third, in order for attribute dimensions to be effectively employed in the model, it must be assumed that there are no interactive effects of the attributes (an assumption made by virtually all multiattribute models). If an interaction occurs, the ability of the model to predict satisfaction will be diminished. Nonetheless, while there may be situations where this will occur, we would argue that the proposed model can account for a large proportion of the variability in consumer satisfaction. In addition, should the need arise, interaction terms can be included in the model.

Finally, we are not suggesting that this model be employed as a primary measure of consumer satisfaction for various products and services. Simple, direct measures of satisfaction are available for this purpose. What we have proposed is a conceptual model of the major determinants of consumer satisfaction which, if verified by empirical research, should provide further insight into the causes of consumer satisfaction/dissatisfaction.

CONCLUSIONS

Our analysis of consumer satisfaction research suggests that close attention to conceptual and methodological issues will be necessary if further progress is to be made toward understanding the determinants of consumer satisfaction. Hopefully, the proposed psychological model of satisfaction will contribute important insight into these determinants.

REFERENCES

Olli T. Ahtola, "The Vector Model of Preferences: An Alternative to the Fishbein Model," Journal of Marketing Research, 12 (February, 1975), 52-9.

Rolph E. Anderson, "Consumer Dissatisfaction: The Effect of Disconfirmed Expectancy on Perceived Product Performance," Journal of Marketing Research, 10(February, 1973), 38-44.

Stephen B. Ash, "A Comprehensive Study of Consumer Satisfaction with Durable Products," in H. Keith Hunt, ed., Advances in Consumer Research, Volume V, (Ann Arbor: Association for Consumer Research, 1978), 254-62.

Richard N. Cardozo, "An Experimental Study of Consumer Effort, Expectation and Satisfaction," Journal of Marketing Research, 2 (August, 1965), 244-9.

J. Merrill Carlsmith and Elliot Aronson, "Some Hedonic Consequences of the Confirmation and Disconfirmation of Expectancies," Journal of Abnormal and Social Psychology, 66 (February, 1963), 151-6.

Ralph L. Day and Muzafer Bodur, "A Comparative Study of Satisfaction with Consumer Services and Intangibles," in Ralph L. Day, ed., Consumer Satisfaction, Dissatisfaction, and Complaining Behavior, Division of Business Research, Indiana University, 1977, 64-74.

James F. Engel, David T. Kollat and Roger D. Blackwell, Consumer Behavior (second ed.), (New York: Holt, Rinehart and Winston, 1973).

Leon Festinger, A Theory of Cognitive Dissonance, (Evanston, Illinois: Row, Peterson, 1957).

Martin Fishbein, "A Behavior Theory Approach to the Relations Between Beliefs About an Object and the Attitude Toward the Object," in Martin Fishbein, ed.) Readings in Attitude Theory and Measurement, (New York: John Wiley & Sons, 1967).

Martin Fishbein and Icek Ajzen, Belief, Attitude, Intention and Behavior, (Reading, Massachusetts: Addison-Wesley Publishing Company, 1975).

Kjell Gr°nhaug, "Exploring Consumer Complaining Behavior: A Model and Some Empirical Results," in W. D. Perreault, ed., Advances in Consumer Research, Volume IV, (Atlanta: Association for Consumer Research, 1977), 159-165.

Charles R. Handy, "Monitoring Consumer Satisfaction with Food Products," in H. Keith Hunt, ed., Conceptualization and Measurement of Consumer Satisfaction and Dissatisfaction, (Cambridge, Massachusetts: Marketing Science Institute, 1977).

Charles R. Handy and Martin Pfaff, "Consumer Satisfaction with Food Products and Marketing Services," Agricultural Economic Report No. 281, Economic Research Service, U.S. Dept. of Agriculture, 1975.

Harry Helson, Adaptation-Level Theory, (New York: Harper and Row, 1964).

John A. Howard and Jagdish N. Sheth, The Theory of Buyer Behavior (New York: Wiley, 1969).

Harold H. Kassarjian, "Personality and Consumer Behavior: A Review," Journal of Marketing Research, 8 (November, 1971), 409-18.

Philip Kotler, Marketing Management: Analysis, Planning, and Control, (Englewood Cliffs, NJ: Prentice-Hall, 1976).

John A. Miller, "Studying Satisfaction, Modifying Models, Eliciting Expectations, Posing Problems, and Making Meaningful Measurements," in H. Keith Hunt, ed., Conceptualization and Measurement of Consumer Satisfaction and Dissatisfaction, (Cambridge, Massachusetts: Marketing Science Institute, 1977).

Richard N. Olshavsky and John A. Miller, "Consumer Expectations, Product Performance, and Perceived Product Quality,'' Journal of Marketing Research, 9 (February, 1972), 19-21.

Jerry C. Olson and Philip Dover, "Effects of Expectation Creation and Disconfirmation on Belief Elements of Cognitive Structure," in B. B. Anderson, ed., Advances in Consumer Research, Volume III, (Chicago: Association for Consumer Research, 1976), 168-75.

Martin Orne, "On the Social Psychology of the Psychological Experiment: With Particular Reference to Demand Characteristics and Their Implications," American Psychologist, 17 (1962), 776-83.

M. J. Rosenberg, "The Conditions and Consequences of Evaluation Apprehension," in R. Rosenthal and R. L. Rosnow, eds., Artifacts in Behavioral Research, (New York: Academic Press, 1969).

Muzafer Sherif and Carl I. Hovland, Social Judgements: Assimilation and Contrast Effects in Communication and Attitude Change, (New Haven, Connecticut: Yale University Press, 1961).

John O. Summers and Donald H. Granbois, "Predictive and Normative Expectations in Consumer Satisfaction and Complaining Behavior," in W. D. Perreault, ed., Advances in Consumer Research, Volume IV, (Atlanta: Association for Consumer Research, 1977), 155-8.

John E. Swan and Linda Jones Combs, "Product Performance and Consumer Dissatisfaction: A New Concept," Journal of Marketing Research, 40 (April, 1976), 25-33.

J. W. Thibaut and H. H. Kelley, The Social Psychology of Groups, (New York: Wiley, 1959).

----------------------------------------